{ "run_id": "3cac5ed2b7bd", "note": "Measured on the device described in `hardware`. Synthetic Gaussian inputs of the model input shapes; batch 1; single process; paced open-loop load. This is inference latency and throughput, not capture or flow-extraction performance.", "hardware": { "host": "jetsonthor", "machine": "aarch64", "kernel": "6.8.12-tegra", "soc": "tegra264", "devicetree_compatible": [ "nvidia,p4071-0000+p3834-0008", "nvidia,p3834-0008", "nvidia,tegra264" ], "is_jetson": true, "l4t_release": "# R38 (release), REVISION: 4.0, GCID: 43443517, BOARD: generic, EABI: aarch64, DATE: Wed Dec 31 00:15:19 UTC 2025", "nvpmodel": "NV Power Mode: 120W 1", "cpu_count": 14, "mem_total_gb": 122.82, "loadavg_1m_at_start": 3.146484375, "loadavg_5m_at_start": 2.60302734375, "python": "3.12.3", "onnxruntime": "1.29.0", "numpy": "2.5.3", "available_providers": [ "CUDAExecutionProvider", "CPUExecutionProvider" ], "git_sha": null, "benchmark_started_at": "2026-09-08T18:43:58.796942+00:00" }, "benchmark_finished_at": "2026-09-08T19:15:59.514576+00:00", "duration_per_tier_s": 300.0, "tiers_requested_rps": [ 10.0, 100.0, 1000.0 ], "models": [ { "model": "detector", "onnx": "gbm_detector.onnx", "onnx_sha256_16": "dfd424e674becf63", "provider": "CPUExecutionProvider", "provider_error": "Fail: [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running Slice node. Name:'' Status Message: CUDA error cudaErrorNoKernelImageForDevice:no kernel image is available for execution on the device", "provider_requested": "auto", "trt_requested": false, "single_inference_p50_ms": 0.0211, "tiers": [ { "n_samples": 3001, "elapsed_s": 300.0, "actual_rps": 10.0, "deadline_misses": 0, "p50_ms": 0.0296, "p95_ms": 0.0596, "p99_ms": 0.1697, "min_ms": 0.021, "max_ms": 0.2952, "mean_ms": 0.0358, "target_rps": 10.0, "provider": "CPUExecutionProvider", "process_rss_gb": 0.3443, "tegrastats": { "n_samples": 299, "rails_mw": { "VDD_GPU": { "p50": 1965, "peak": 3144, "min": 1965 }, "VDD_CPU_SOC_MSS": { "p50": 7862, "peak": 9434, "min": 7469 }, "VIN_SYS_5V0": { "p50": 7608, "peak": 8128, "min": 7608 }, "VIN": { "p50": 24440, "peak": 27962, "min": 22842 } }, "temps_c": { "cpu": { "p50": 38.125, "peak": 38.687 }, "tj": { "p50": 39.218, "peak": 39.625 }, "soc012": { "p50": 37.687, "peak": 38.218 }, "gpu": { "p50": 38.312, "peak": 38.687 }, "soc345": { "p50": 39.218, "peak": 39.625 } }, "board_ram_used_mb": { "p50": 36776, "peak": 36779 } } }, { "n_samples": 30001, "elapsed_s": 300.003, "actual_rps": 100.0, "deadline_misses": 0, "p50_ms": 0.025, "p95_ms": 0.0559, "p99_ms": 0.1067, "min_ms": 0.0091, "max_ms": 8.5565, "mean_ms": 0.0448, "target_rps": 100.0, "provider": "CPUExecutionProvider", "process_rss_gb": 0.3457, "tegrastats": { "n_samples": 298, "rails_mw": { "VDD_GPU": { "p50": 1964, "peak": 1965, "min": 1964 }, "VDD_CPU_SOC_MSS": { "p50": 34580, "peak": 35366, "min": 27518 }, "VIN_SYS_5V0": { "p50": 7912, "peak": 7912, "min": 7709 }, "VIN": { "p50": 53572, "peak": 55208, "min": 47022 } }, "temps_c": { "cpu": { "p50": 54.062, "peak": 56.75 }, "tj": { "p50": 54.093, "peak": 56.75 }, "soc012": { "p50": 52.25, "peak": 54.812 }, "gpu": { "p50": 46.187, "peak": 48.5 }, "soc345": { "p50": 49.312, "peak": 51.625 } }, "board_ram_used_mb": { "p50": 36773, "peak": 36779 } } }, { "n_samples": 300001, "elapsed_s": 300.0, "actual_rps": 1000.0, "deadline_misses": 43118, "p50_ms": 0.0213, "p95_ms": 0.0237, "p99_ms": 0.028, "min_ms": 0.0081, "max_ms": 14.7371, "mean_ms": 0.0227, "target_rps": 1000.0, "provider": "CPUExecutionProvider", "process_rss_gb": 0.3558, "tegrastats": { "n_samples": 299, "rails_mw": { "VDD_GPU": { "p50": 1964, "peak": 2357, "min": 1964 }, "VDD_CPU_SOC_MSS": { "p50": 35744, "peak": 35759, "min": 34173 }, "VIN_SYS_5V0": { "p50": 7912, "peak": 8013, "min": 7912 }, "VIN": { "p50": 54298, "peak": 56054, "min": 52754 } }, "temps_c": { "cpu": { "p50": 58.187, "peak": 59.562 }, "tj": { "p50": 58.187, "peak": 59.562 }, "soc012": { "p50": 56.125, "peak": 57.375 }, "gpu": { "p50": 50.062, "peak": 51.125 }, "soc345": { "p50": 53.156, "peak": 54.093 } }, "board_ram_used_mb": { "p50": 36771, "peak": 36774 } } } ] }, { "model": "forecaster", "onnx": "ar_forecaster.onnx", "onnx_sha256_16": "169ca3f7290decfb", "provider": "CPUExecutionProvider", "provider_error": "Fail: [ONNXRuntimeError] : 1 : FAIL : Non-zero status code returned while running Slice node. Name:'' Status Message: CUDA error cudaErrorNoKernelImageForDevice:no kernel image is available for execution on the device", "provider_requested": "auto", "trt_requested": false, "single_inference_p50_ms": 0.0134, "tiers": [ { "n_samples": 3001, "elapsed_s": 300.0, "actual_rps": 10.0, "deadline_misses": 0, "p50_ms": 0.2218, "p95_ms": 0.2246, "p99_ms": 0.2264, "min_ms": 0.0151, "max_ms": 1.6224, "mean_ms": 0.2202, "target_rps": 10.0, "provider": "CPUExecutionProvider", "process_rss_gb": 0.3585, "tegrastats": { "n_samples": 299, "rails_mw": { "VDD_GPU": { "p50": 1965, "peak": 1965, "min": 1964 }, "VDD_CPU_SOC_MSS": { "p50": 7469, "peak": 24756, "min": 7469 }, "VIN_SYS_5V0": { "p50": 7709, "peak": 8128, "min": 7709 }, "VIN": { "p50": 24458, "peak": 26018, "min": 23048 } }, "temps_c": { "cpu": { "p50": 42.75, "peak": 51.406 }, "tj": { "p50": 44.125, "peak": 52.593 }, "soc012": { "p50": 42.906, "peak": 51.5 }, "gpu": { "p50": 43.031, "peak": 51.0 }, "soc345": { "p50": 44.125, "peak": 52.593 } }, "board_ram_used_mb": { "p50": 36772, "peak": 36774 } } }, { "n_samples": 30001, "elapsed_s": 300.0, "actual_rps": 100.0, "deadline_misses": 0, "p50_ms": 0.014, "p95_ms": 0.0147, "p99_ms": 0.0164, "min_ms": 0.0134, "max_ms": 0.3372, "mean_ms": 0.0142, "target_rps": 100.0, "provider": "CPUExecutionProvider", "process_rss_gb": 0.3585, "tegrastats": { "n_samples": 300, "rails_mw": { "VDD_GPU": { "p50": 1965, "peak": 1965, "min": 1965 }, "VDD_CPU_SOC_MSS": { "p50": 7469, "peak": 8255, "min": 7469 }, "VIN_SYS_5V0": { "p50": 7709, "peak": 8026, "min": 7608 }, "VIN": { "p50": 24342, "peak": 26664, "min": 23168 } }, "temps_c": { "cpu": { "p50": 39.968, "peak": 40.937 }, "tj": { "p50": 41.218, "peak": 42.281 }, "soc012": { "p50": 40.031, "peak": 41.062 }, "gpu": { "p50": 40.125, "peak": 41.156 }, "soc345": { "p50": 41.218, "peak": 42.281 } }, "board_ram_used_mb": { "p50": 36776, "peak": 36777 } } }, { "n_samples": 300001, "elapsed_s": 300.0, "actual_rps": 1000.0, "deadline_misses": 0, "p50_ms": 0.0138, "p95_ms": 0.0141, "p99_ms": 0.0143, "min_ms": 0.0132, "max_ms": 1.0395, "mean_ms": 0.0138, "target_rps": 1000.0, "provider": "CPUExecutionProvider", "process_rss_gb": 0.3639, "tegrastats": { "n_samples": 299, "rails_mw": { "VDD_GPU": { "p50": 1965, "peak": 1965, "min": 1965 }, "VDD_CPU_SOC_MSS": { "p50": 7469, "peak": 8255, "min": 7469 }, "VIN_SYS_5V0": { "p50": 7608, "peak": 8026, "min": 7608 }, "VIN": { "p50": 24424, "peak": 26194, "min": 23152 } }, "temps_c": { "cpu": { "p50": 39.062, "peak": 39.437 }, "tj": { "p50": 40.25, "peak": 40.687 }, "soc012": { "p50": 39.062, "peak": 39.468 }, "gpu": { "p50": 39.218, "peak": 39.656 }, "soc345": { "p50": 40.25, "peak": 40.687 } }, "board_ram_used_mb": { "p50": 36774, "peak": 36776 } } } ] } ], "memory": { "process_peak_rss_gb": 0.3639, "process_rss_before_gb": 0.0552, "process_rss_after_gb": 0.3639, "note": "Process RSS only. Board RAM is in the tegrastats summaries when present." }, "idle_baseline": { "seconds": 60, "before_load": { "n_samples": 59, "rails_mw": { "VDD_GPU": { "p50": 1965, "peak": 1965, "min": 1965 }, "VDD_CPU_SOC_MSS": { "p50": 7469, "peak": 7472, "min": 7469 }, "VIN_SYS_5V0": { "p50": 7709, "peak": 7709, "min": 7709 }, "VIN": { "p50": 24170, "peak": 25164, "min": 23108 } }, "temps_c": { "cpu": { "p50": 38.125, "peak": 38.281 }, "tj": { "p50": 39.25, "peak": 39.375 }, "soc012": { "p50": 37.781, "peak": 37.968 }, "gpu": { "p50": 38.468, "peak": 38.656 }, "soc345": { "p50": 39.25, "peak": 39.375 } }, "board_ram_used_mb": { "p50": 36621, "peak": 36621 } }, "after_load": { "n_samples": 60, "rails_mw": { "VDD_GPU": { "p50": 1965, "peak": 1965, "min": 1965 }, "VDD_CPU_SOC_MSS": { "p50": 7469, "peak": 7862, "min": 7469 }, "VIN_SYS_5V0": { "p50": 7608, "peak": 8026, "min": 7608 }, "VIN": { "p50": 24146, "peak": 25642, "min": 22884 } }, "temps_c": { "cpu": { "p50": 38.75, "peak": 38.906 }, "tj": { "p50": 39.937, "peak": 40.062 }, "soc012": { "p50": 38.718, "peak": 38.875 }, "gpu": { "p50": 38.937, "peak": 39.031 }, "soc345": { "p50": 39.937, "peak": 40.062 } }, "board_ram_used_mb": { "p50": 36776, "peak": 36776 } }, "note": "Board rails and temperatures while this process slept; other host activity is included." }, "tegrastats_samples_file": "thor_benchmark_tegrastats.jsonl", "tegrastats_sample_count": 1914, "gates": { "detector_p95_latency_ms": { "threshold": 10.0, "unit": "ms", "measured": 0.0237, "status": "pass", "criterion": "p95 single-event latency at the 1000 events/s tier" }, "forecaster_p95_latency_ms": { "threshold": 50.0, "unit": "ms", "measured": 0.0141, "status": "pass", "criterion": "p95 single-sequence latency at the 1000 events/s tier" }, "throughput_at_1000_rps": { "threshold": 1000.0, "unit": "events/s", "measured": 1000.0, "status": "pass", "criterion": "lowest achieved events/s across models at the 1000 events/s tier; pass at >= 99% of target" }, "memory_footprint_gb": { "threshold": 4.0, "unit": "GB", "measured": 0.3639, "status": "pass", "criterion": "peak resident set size of the benchmark process (ru_maxrss)" } }, "source_badge": "validated-thor-benchmark" }